1. Introduction

Rapid progress in the development of high-power tabletop terahertz (THz) sources in recent years [1,2] has seen the emergence of a wide variety of new applications of THz radiation, such as in nonlinear spectroscopy [3,4], imaging [5], control over the properties of condensed matter [6–8], and THz-driven particle acceleration [9–13]. Many of the advancements in intense THz sources have centred on the development of broadband, single-cycle THz pulses, owing to their capacity to achieve high peak field strengths [14–17]. However for many applications such as THz-driven particle acceleration [9], THz imaging [5], and control over material properties [7], high-energy narrowband, multi-cycle THz pulses are highly desirable but prove much more challenging to generate with comparable field strengths and laser-to-THz conversion efficiencies.

Lithium niobate (LN) has emerged as a leading material for generating high-energy narrowband THz pulses, with optical techniques such as chirped-pulse beating combined with tilted-pulse-front phase-matching schemes required for bulk LN crystals [9,18,19]. Alternative material systems have been investigated as multi-cycle THz sources, such as semiconductors [20] and organic crystals [21,22], however these sources often suffer from lower damage thresholds or conversion efficiencies. Periodically-poled lithium niobate (PPLN) crystals [23–32] offer a convenient alternative to bulk LN crystals, as PPLN directly generates narrowband THz radiation whilst simultaneously satisfying the challenging phase-matching requirements of LN without requiring any complex pulse-front tilting techniques. Ultra-narrowband (relative bandwidth of $<1\%$) THz pulses with energies up to 458 $\mu$J at 0.361 THz have been reported using a cryogenically-cooled PPLN crystal [31], and recently internal conversion efficiencies of up to 0.9 % have been demonstrated for cryogenically-cooled PPLN crystals pumped by a two-line laser system [32]. Despite the benefits of PPLN, the requirement of strong electric fields to overcome the coercive field of LN during fabrication typically restricts the crystal size to an aperture of $<1$ cm$^2$, limiting their scalability to higher energy pump laser systems.

A solution to these scalability issues was demonstrated by Lemery et al. [33], who fabricated a PPLN structure by stacking individual large-area ($\approx 50$ mm diameter) LN wafers together by hand in air, with the ferroelectric $Z$-axis of each consecutive wafer rotated by 180$^{\circ }$ to satisfy quasi-phase matching. This hand-built PPLN wafer stack source was able to generate 1.3 mJ narrowband THz pulses at a frequency of 0.16 THz when driven by a Joule-class laser system, exploiting well-established wafer fabrication techniques to demonstrate large-area, low-cost and highly-scalable narrowband THz sources, without resorting to complex wafer-bonding techniques. However, a detailed understanding of the physical mechanisms driving the PPLN wafer-stack source performance are required to achieve further key developments, including higher-frequency operation, frequency tunability, and improved conversion efficiency, which are essential to access the full potential of these novel narrowband THz sources.

Here we address this through detailed investigation, exploring PPLN wafer-stack sources with thinner 135 $\mu$m-thick wafers to produce higher frequencies (up to 0.39 THz) directly suitable for THz-driven particle acceleration applications. The fluence and intensity dependence of the THz conversion efficiency and pulse energy are examined through spatio-temporal variations of the optical pump pulse, with a peak efficiency of 0.17% achieved for 750 fs chirped pulses under 700 mJ/cm$^{2}$ excitation conditions. Electro-optic sampling measurements of the multi-cycle THz fields allow us to study pulse build-up with wafer number, leading to the observation and explanation of unexpected independent frequency shifts (up to 5%), due to THz etalon effects in the inter-wafer air gaps, and the influence of the THz Gouy phase on the phase-matching condition arising from narrow transverse dimension beams. These findings are critical to develop optimized high-energy narrowband THz sources at higher, precisely-defined frequencies that will drive future novel narrow-bandwidth applications.

2. Theory

In a periodically-poled structure, the length scale on which the sign of the nonlinear coefficient is inverted is designed to match the coherent build-up length of the nonlinear interaction. For THz generation in PPLN this length scale is determined by the walk-off between the THz and pump pulses as they propagate through LN. The central frequency of THz radiation produced by a PPLN structure can therefore be calculated as

The wafers in the stack are not optically bonded, which leaves open the potential for both optical and THz reflection losses at each wafer boundary. To remove the optical losses, all wafers are anti-reflection coated at $760 - 840$ nm. As noted by Lemery et al. [33], small $\leq 10\,\mu$m gaps are readily obtained by physical adhesive-free stacking, and permit high power transmission at THz frequencies between each wafer pair (99.6% for 1 $\mu$m gap, 96% for 10 $\mu$m gap at 0.39 THz). However, as we show in section 4.3, the phase-shift of such sub-wavelength inter-wafer gaps is not negligible, and a significant frequency tuning arises from the inclusion of the wafer-wafer phase-shift into the phase-matching condition.

A 1-dimensional model of THz generation in PPLN wafer stacks, in the absence of inter-wafer phase shifts, gives the total THz electric field at the exit surface of the PPLN stack from the sum of contributions from individual wafers,

 

Fig. 1. (a) Simulated power spectra of THz pulses generated from a 10-wafer PPLN stack consisting of $135\,\mu$m-thick wafers, using 40 fs and 750 fs duration laser pulses (red and black lines, respectively). Inset are the time-domain waveforms (offset vertically for clarity), where the dashed lines are THz pulses generated from a single wafer, for comparison. (b) Schematic diagram showing the front view of a single wafer. (c) Schematic diagram of the side view of a 10-wafer PPLN stack, where the red arrows denote the direction of the $+Z$ axis in each wafer.

Download Full Size | PDF

The simulations of Figure 1 assume no gaps between wafers; the effects of a finite gap between wafers will be addressed in Section 4.3. The extension to 3-dimensional optical beams and the inclusion of Gouy phase-shifts also results in a frequency tuning, as discussed in Section 4.4.

3. Methods

Single crystal wafers of congruent lithium niobate are commercially available, with a diameter of $50.8\pm 0.2$ mm and a thickness of $135\pm 20\,\mu$m. We received two batches of wafers, which were found to have different thicknesses: batch 1 had an average thickness of 135 $\mu$m, and batch 2 had an average thickness of 155 $\mu$m. All wafers had a surface orientation of $X\pm \,0.2^{\circ }$, with the $Z$-axis indicated as perpendicular to the wafer flat, as shown in Fig. 1(b). Wafer stacks were created in air by hand, with periodic poling achieved by inverting the direction of the +$Z$ crystal axis in adjacent wafers, as shown in Fig. 1(c). Stacking was performed by laterally sliding each wafer on top of the previous one; this not only helps minimize gaps between wafers by removing air pockets, but also achieves good adhesion between wafers. Wafers could subsequently be removed from the stack by sliding, but could not be pulled apart by a force applied perpendicular to the surface. The wafer total thickness variation (TTV) was less than 10 $\mu$m, facilitating small inter-wafer gaps and high THz transmission between wafers.

The laser used in this work was a regeneratively-amplified Ti:Sapphire system, which produced 6 mJ pulses with a repetition rate of 1 kHz, 800 nm central wavelength, and a transform-limited pulse duration of 40 fs full-width at half-maximum (FWHM). The pump pulse energy at the surface of the PPLN wafer stack was 5 mJ after reflection losses during transport, resulting in a fluence of 10 mJ/cm$^{2}$ from a collimated pump beam. The THz and pump beams were decoupled using a 1 mm-thick PTFE plate positioned behind the PPLN wafer stack, blocking the pump laser whilst maintaining high THz transmission. The generated THz radiation was collected, collimated, and then focused for detection by a pair of $50.8$ mm focal length off-axis parabolic mirrors. The pump fluence on the PPLN stack was varied by introducing a translatable lens into the pump beam, allowing the spot size of the pump laser on the PPLN wafer stack to be varied. In this case, the THz and pump beams were decoupled by focusing the pump beam through a hole in the centre of the first parabolic mirror, with the PTFE plate positioned in between the two parabolic mirrors to remove any residual pump radiation. Laser pulse duration and chirp were varied using the internal compressor of the regenerative amplifier, with grating separation in the compressor calibrated to FWHM pulse duration by autocorrelation measurements. A small portion of the laser beam was split off to be used as a probe for electro-optic (EO) detection measurements, performed using a (110)-oriented, 2 mm-thick ZnTe crystal. THz pulse energy measurements were performed using a calibrated pyroelectric detector (Gentec THZ5I-BL-BNC) positioned at the same focus as the ZnTe crystal. During these pulse energy measurements the EO detection probe beam was redirected onto a beam dump. The manufacturer calibration of the pyroelectric detector was found to agree within $0.2\,\mu$J of our own calibration using a Thomas Keating THz power meter.

4. Results

4.1 Terahertz generation from PPLN stacks with varying numbers of wafers

Assembling custom stacks of PPLN wafers offers a unique opportunity to explore how the narrowband multi-cycle THz pulses build up with increasing numbers of periodically-poled domains, as shown in Fig. 2 by EO sampling measurements of stacks with 4 to 20 wafers. The duration of both pump and probe pulses was 750 fs, and the pump beam was collimated giving a fluence of 10 mJ/cm$^2$. Here, the wafer stacks were assembled such that each additional wafer became the new pump entrance surface of the stack, adding an additional half-cycle to the end of the generated THz waveform, due to the pump pulse travelling faster than the THz pulse through LN. As shown in Fig. 2, each pair of wafers adds an extra full cycle to the THz pulse, providing the unique ability to precisely control the bandwidth of the PPLN source. The electric field amplitude at a given time delay remains almost constant as the number of wafers in the stack increases, with only a small decrease in field amplitude observed in later cycles relative to earlier cycles due to THz absorption losses in LN.

 

Fig. 2. THz time-domain waveforms produced by PPLN wafer stacks as the number of wafers in the stack was increased from 4 to 20, with a pump fluence of 10 mJ/cm$^{2}$ and pulse duration of 750 fs. Waveforms are offset vertically for clarity and given a time offset to line up the contribution to the pulse from the exit wafer of the stack, indicated by the vertical dashed line.

Download Full Size | PDF

Figure 3 shows the power spectrum of the THz pulse produced by the 20-wafer PPLN stack, with the corresponding measured time-domain waveform in the inset. Here the PPLN wafer stack was mounted between two UV-grade fused silica (UVFS) windows (2"-diameter, 12 mm-thick, double-side anti-reflection coated for $650-1050$ nm) to minimize the influence of internal THz reflections on the measured waveform and allow measurement over a wider time-domain window. The central frequency of the spectrum occurs at 0.39 THz, which is in good agreement with the prediction from Eq. (1) for 135 $\mu$m-thick wafers. The 8 GHz discrepancy between the theoretical (0.398 THz) and experimentally-observed central frequency can be explained by the influence of inter-wafer gaps on the THz phase-matching condition, which we explore in section 4.3. The main THz pulse is visible in the inset of Fig. 3 from 0 - 25 ps, while also visible is a longer wavelength pulse from 40 - 90 ps, which arises due to THz radiation generated propagating in the opposite direction to the pump beam. This backwards-propagating radiation is subject to a different coherent build-up length, hence has a different frequency given by $c/\left [\Lambda \left (n_{\mathrm {THz}}+n_{\mathrm {IR}}\right )\right ]=0.152$ THz, in good agreement with a small peak observed in the experimental spectrum.

 

Fig. 3. Power spectrum of the THz pulse produced by a PPLN wafer stack consisting of 20 wafers, each of 135-$\mu$m thickness. Inset is the time-domain waveform of the THz pulse.

Download Full Size | PDF

These measurements highlight the versatility of PPLN wafer-stack THz sources, providing the ability to structure THz pulses with any desired pulse duration and bandwidth by simply choosing the number of wafers in the stack. This is in contrast to conventional PPLN THz sources which are limited to a fixed number of poled domains by the manufacturing process.

4.2 Optimizing high-energy terahertz generation from PPLN wafer stacks

We explored the performance of our wafer stacks under intense laser conditions, optimizing the fluence, duration, and chirp of the laser pump pulse to achieve maximum laser-to-THz conversion efficiency. LN is susceptible to optical damage at high laser fluence, for example a laser-induced damage threshold of 400 mJ/cm$^{2}$ was observed using 40 fs pulses from a 1 kHz laser [37], although damage threshold increases with increasing pulse duration. To avoid damaging the 135 $\mu$m-thick wafers from batch 1, another PPLN wafer stack was created for these measurements, consisting of 10 wafers of 155 $\mu$m-thickness from batch 2, which produced narrowband THz pulses with a central frequency of 0.35 THz.

Figure 4(a) presents the THz conversion efficiency at different pump fluences, as a function of pump pulse duration and chirp. At each fluence, the THz conversion efficiency initially exhibits a linear increase with decreasing pump pulse duration, before saturating and then decreasing towards the shortest pulses. For a given pulse duration, the magnitude of the THz conversion efficiency is observed to increase with increasing pump fluence up to 700 mJ/cm$^{2}$, and decrease at higher fluence. The peak in conversion efficiency is observed to shift to a longer pulse duration for higher fluence, with the maximum conversion efficiency of 0.17% obtained at a fluence of 700 mJ/cm$^{2}$ and pulse duration of 750 fs. A similar trend is observed when using positively and negatively chirped pulses at a fluence of 100 mJ/cm$^{2}$; however, when the fluence is increased to 700 mJ/cm$^{2}$, THz conversion efficiency is on average $0.013$ percentage points higher with positively chirped pulses compared to negatively chirped pulses of the same duration (corresponding to an $8\%$ increase in maximum conversion efficiency). Figure 4(b) presents THz conversion efficiency as a function of laser intensity, for various pump pulse durations. At low intensity, THz conversion efficiency increases linearly with pump intensity, as expected for an optical rectification process, whilst deviation from this linear trend is observed above an intensity of approximately 0.4 TW/cm$^{2}$.

 

Fig. 4. (a) THz conversion efficiency for various pump fluences as a function of chirped pump pulse duration, generated from a PPLN wafer stack consisting of 10 wafers each of 155 $\mu$m thickness. (b) THz conversion efficiency as a function of pump intensity for various pulse durations. The red dashed line is a guide to the eye highlighting the linear trend at low intensity. (c) Pump laser spectrum before and after propagating through the PPLN wafer stack with a fluence of 900 mJ/cm$^2$ and pulse duration of 750 fs.

Download Full Size | PDF

We observe a significant redshift in the spectrum of high intensity pump pulses after propagation through the PPLN wafer stack, as shown in Fig. 4(c) for a 10-wafer PPLN stack with an incident intensity on the order of $1.2$ TW/cm$^{2}$. Although a red-shifted pump spectrum can correspond to cascaded high-efficiency THz generation beyond the Manley-Rowe limit (as discussed by Lemery et al. [33] for their wafer stacks), we observe a greater spectral redshift for higher intensity pulses but no clear correlation with the THz conversion efficiency behavior shown in Figure 4 (a). Hence we attribute the spectral redshift and the decrease in THz conversion efficiency for shorter, more intense pump pulses to parasitic nonlinear effects degrading the laser pulse as it propagates through the PPLN wafer stack. Additionally, as shown in Figure 4(c), the pump pulse spectrum has broadened to wavelengths slightly outside the range of the wafer AR coating, which may result in increased reflection losses at each wafer-air interface.

We also investigated how the THz pulse energy changed as a function of the number of wafers in the PPLN stack, repeating the EO sampling measurements from Fig. 2 with a higher pump fluence of 530 mJ/cm$^2$. As before, 750 fs pump and probe pulses were used, putting this close to the peak conversion efficiency observed in Fig. 4. An estimate of the THz pulse energy was extracted from the time-domain waveforms using the well-known equations relating the peak electric field, peak irradiance and total power of a Gaussian beam, $I_{0}=E_\mathrm {0}/2\eta _{0}$ and $I_{0}=2P/\pi {w_{0}^2}$, where $\eta _{0}$ is the impedance of free space and $w_0$ is the beam waist. Pulse energy was then calculated as $W_\mathrm {THz}=P/\tau$ using the measured pulse duration $\tau$. This data is presented in Fig. 5, demonstrating a linear increase in pulse energy at low pump fluence, whilst at high fluence the pulse energy is observed to saturate with increasing numbers of wafers in the stack. Data in Fig. 5 are calibrated to an energy measurement by a pyroelectric detector using the 20-wafer stack (circled data point). Insight into the mechanism behind this saturation can be gained by comparing the time-domain waveforms from the high and low fluence case. The inset of Fig. 5 shows THz pulses produced by 4- and 20- wafer stacks at high fluence, where a factor of 2 decrease in electric field amplitude is observed when comparing the first two cycles of the pulse from the 20-wafer stack to the pulse from the 4-wafer stack, despite the THz radiation being generated from the same wafers. This is in contrast to the low fluence case shown in Fig. 2 where the amplitude remained similar regardless of wafer number. This is a further indication for the influence of parasitic nonlinear effects degrading the laser pulse in the saturation of THz conversion efficiency, and highlights the importance of understanding the role of such nonlinear effects in pushing THz generation from PPLN past its current limits.

 

Fig. 5. Energy of THz pulses produced by PPLN wafer stacks consisting of 4 to 20 wafers, at high and low pump fluence. Data are extracted from time-domain waveforms and calibrated by a pyroelectric detector measurement with the 20-wafer PPLN stack (circled data point). Error bars show a 95 % confidence interval. Dashed lines are fits to the data to highlight the trends. Inset are the time-domain waveforms generated by 4- and 20-wafer PPLN stacks at 530 mJ/cm$^{2}$ pump fluence.

Download Full Size | PDF

4.3 Effect of inter-wafer gaps on THz pulse frequency

We observe a significant shift in the central frequency of THz pulses produced by the PPLN wafer stacks depending on the mounting method used. Figure 6(a) presents time-domain waveforms of THz pulses obtained first with a 20-wafer PPLN stack mounted between two apertured plates applying light pressure to the outer edge ($\approx 1$ mm) of the wafer surfaces, and then the same PPLN wafer stack compressed between two 2"-diameter UVFS windows applying uniform pressure across the whole surface of the wafers. In both cases the pump beam was collimated with a fluence of 10 mJ/cm$^{2}$. A reduction in the period of the THz pulse is observed when compressing the stack between the UVFS windows compared to mounting by the edge of the wafers, with the power spectra in Fig. 6(b) demonstrating a 10 GHz shift to higher central frequency.

 

Fig. 6. (a) Experimentally measured time-domain waveforms of the THz pulse generated by a 20-wafer PPLN stack, with (red) and without (blue) the stack compressed between two UVFS windows. Solid and dashed black lines are simulated THz pulses matching the traces with and without the stack compressed between UVFS windows, respectively. (b) Fourier transform spectra of the corresponding color waveforms in panel (a). (c) Simulated absolute value and phase of the amplitude transmission coefficient of a single wafer gap etalon, as a function of gap size, for 0.398 THz radiation. (d) Central frequency of simulated THz pulses generated by 20-wafer PPLN stacks with varying gap size between wafers.

Download Full Size | PDF

The frequency shift can be explained by accounting for the air gap between wafers in the PPLN stack acting as a THz Fabry-Pérot etalon. While the power transmission is near unity, the electric field amplitude transmission coefficient ($t_{\mathrm {e}}$) of the LN-air-LN gap can give rise to a significant additional phase-shift ($\psi$) where,

Here $t$ and $r$ are the THz field transmission and reflection coefficients of the LN-air interface, respectively, $\delta =4\pi {n_{\mathrm {THz}}d}/\lambda _\mathrm {THz}$ and $d$ is the gap size. Adding this phase-shift to the optical-THz group-phase slippage in LN is sufficient to cause a significant modification to the PPLN phase-matching condition and generated THz frequency. Experimentally, compressing the wafer stack between two UVFS windows reduced the gap between wafers, leading to smaller phase-shifts and higher THz frequency, as observed in Fig. 6(b).

The effects of the wafer gap on the THz pulses produced by the PPLN wafer stacks were simulated by applying the effects of Eq. (3) to the contribution from each wafer in Eq. (2),

The results presented in Fig. 6 suggest a possible route to fine-tuning the frequency of THz pulses produced by PPLN wafer stacks, by controlling the gap size between wafers. Tunability on the order of 10 GHz, as experimentally demonstrated here, is essential for applications requiring precise narrowband frequencies, such as particle acceleration in THz-driven waveguides [9]. Whilst frequency-tuning of LN THz sources can be achieved without reflection losses by cryogenic cooling [24,26], fine-tuning using wafer gaps represents a complementary technique that is achievable at room-temperature, and is not available to bulk PPLN sources.

4.4 Effect of THz Gouy phase on pulse frequency

We observed further frequency-shifting behavior related to the transverse size of the pump beam, independent of the effect from inter-wafer gaps. Figures 7(a) and 7(b) present the time-domain waveform and the power spectrum, respectively, of the THz pulse generated from a 20-wafer PPLN stack where the focusing pump beam had a 1.1 mm FWHM diameter at the stack (530 mJ/cm$^{2}$ fluence). The central frequency of this THz pulse was 0.36 THz, 15 GHz lower than the 0.375 THz central frequency pulse obtained from the same wafer-stack using an 8 mm FWHM collimated pump beam (10 mJ/cm$^{2}$ fluence). In these measurements, the stack was mounted using the apertured plates and no UVFS substrates, to avoid any effects of a high-fluence pump beam passing through the substrates. From Fig. 6(d), a frequency shift from 0.398 THz to 0.36 THz could have occurred in a 20-wafer stack if the inter-wafer gaps were on the order of $25\,\mu$m; however for a gap of this size $\left |t_{e}\right |=0.89$ and the difference in amplitude between the first and last cycles of the THz pulse would be $0.89^{19}=0.11$. A reduction in amplitude of approximately $0.7$ was observed in Fig. 7(a), hence this frequency shift cannot be explained solely by the influence of inter-wafer gaps.

 

Fig. 7. (a) Time-domain waveforms of THz pulses generated by a 20-wafer PPLN stack using a collimated pump beam (blue squares) and a focusing pump beam (red dots), alongside simulated THz pulses with (solid black curve) and without (dashed black curve) taking into account Gouy phase effects. Waveforms have been offset vertically for clarity. (b) Power spectra of the time-domain waveforms in (a). (c) Calculated Gouy phase of THz radiation at the detection crystal as a function of source position in the stack, applied to the simulations in (a). Wafer position is defined relative to the pump beam entrance surface of the stack.

Download Full Size | PDF

To determine the nature of this additional frequency shift, we extend our model of THz generation from PPLN wafer stacks to include 3-dimensional effects arising from the propagation of THz radiation from source to detector. Each individual wafer in the stack was treated as a separate source of THz radiation with a TEM$_{00}$ Gaussian electric field profile; the $1/e^2$ beam-waist radius ($w_{0}$) and curvature of each THz source was calculated from the pump laser intensity profile, with the THz beam waist being a factor of $\sqrt {2}$ smaller than the laser beam waist as $E_\mathrm {THz}\propto {I_\mathrm {IR}}$ for optical rectification. Propagation of THz radiation was modeled by the evolution of the complex beam parameter $q(z)=z+iz_{R}$ (where $z_{R}$ is the Rayleigh length) with ABCD matrix calculations, taking into account the focusing optics and refractive index of the PPLN source and ZnTe detection crystal. Figure 7(c) presents the simulated THz Gouy phase $\phi _{G}=\tan ^{-1}(z/z_{R})$ at the center of the detection crystal as a function of source position in the 20-wafer PPLN stack, using 0.398 THz radiation generated from sources with $w_{0}=0.66$ mm (corresponding to the measured 1.1 mm FWHM of the pump beam at the stack). A greater than $0.8\,\pi$ phase-shift occurs between THz radiation generated in the entrance and exit wafers in the stack, resulting in the observed frequency shift. In comparison, propagation simulations using a collimated pump beam (THz source $w_{0}=4.81$ mm) produced a maximum Gouy phase shift of only 0.08 rad across the pulse, a factor of 32 smaller than the focused beam.

To simulate the effect of the varying Gouy phase, the contribution to the THz pulse from each wafer including inter-wafer gap effects from Eq. (5) is modified by an additional phase term,

These results demonstrate the critical importance the transverse beam size and transport properties can have on the phase-matching condition for optical rectification-based narrowband THz generation, and highlights that the transverse beam size should be a key consideration in optimizing the performance of PPLN sources when scaling up the THz pulse energy using high-energy laser systems.

5. Conclusion

We have performed a detailed investigation of novel periodically-poled THz sources based on the stacking of large-area ultra-thin LN wafers, demonstrating scalability to higher frequencies, up to 0.39 THz. For the first time, direct waveform measurements through EO sampling (together with pulse energy measurements) have enabled systematic study of PPLN sources as they are assembled wafer by wafer, allowing exploration and optimization of THz frequency, bandwidth and conversion efficiency as a function of the number of wafers for varying pump fluence, duration and transverse properties. In particular, our experimental results have revealed critical frequency-dependence on both stack mounting and pump focusing conditions, which we attribute through supporting simulations to inter-wafer air gap etalon and Gouy phase effects, respectively. We also highlight the potential for parasitic nonlinear effects (such as self-phase modulation) limiting the THz generation efficiency, which must be addressed to extract the full potential from these sources.

Underpinned by our extensive source development here, we have recently used our PPLN wafer stacks in conjunction with a terawatt laser system at Daresbury Laboratory [38] to generate narrowband THz pulses with energies in excess of 100 $\mu$J, for novel THz-driven electron acceleration and manipulation experiments at the Compact Linear Accelerator for Research and Applications (CLARA) test facility. This highlights the direct applicability of these sources, which provide a large area for high-power laser exploitation, straightforward collinear implementation (compared to tilted pulse-front LN setups) and versatile frequency and bandwidth control. Our work here provides the essential development and understanding to maximize the performance and potential of PPLN wafer-stacks and establish them as key sources for applications of high-energy narrowband THz pulses in the field.